Background: Atezo (anti–PD-L1) monotherapy is approved for cis-ineligible pts who have locally advanced or mUC with PD-L1–expressing immune cells on ≥ 5% of the tumor area (IC2/3 per VENTANA SP142 IHC assay). The IMvigor130 primary analysis demonstrated a significant PFS benefit with atezo + platinum/gemcitabine (plt/gem) (arm A) vs placebo (pbo) + plt/gem (arm C) as 1L treatment for mUC (Galsky Lancet 2020); at that time, interim OS data for arm A vs C were encouraging but immature. OS with atezo monotherapy (arm B) could not be formally tested, but favorable efficacy was seen in IC2/3 pts. In this exploratory analysis, we assess outcomes by PD-L1 status in cis-ineligible pts. Methods: Pts were randomized 1:1:1 to arms A, B or C (Galsky Lancet 2020). Evaluation of OS (co-primary EP) was performed via a hierarchical fixed sequence procedure: arm A vs C ITT pts; then, arm B vs C ITT and IC2/3 pts. No formal testing was performed in this exploratory subgroup analyses; OS and RECIST 1.1 ORR (per investigator [secondary EP]) were descriptively evaluated. Results: Efficacy data suggested OS and ORR benefit in atezo-treated cis-ineligible IC2/3 pts (Table). In the overall safety population, all-grade treatment-related AEs (TRAEs) had occurred in 60% and 96% of arm B and C pts, respectively; grade 3-4 TRAEs occurred in 15% and 81%, respectively. Biomarker data evaluating PD-L1 biology (assessed by SP142) and associated transcriptome analysis in arms B vs C will be presented. Conclusions: This exploratory analysis of IMvigor130 pts with 1L cis–ineligible IC2/3 mUC provides additional evidence for clinical benefit with single-agent atezo, a well-tolerated treatment with a distinct safety profile vs chemo. Analyses with longer OS follow-up are warranted. Clinical trial information: NCT02807636. Research Sponsor: F. Hoffmann-La Roche Ltd

mFU, 11.8 mo; data cutoff May 31, 2019. NE, not estimable. ^a Stratified per randomization stratification factors: PD-L1 status (ITT only), Bajorin risk factor score/liver mets, investigator’s choice of plt (cis or carboplatin); Galsky Lancet 2020. ^b Per Galsky Lancet Oncol 2011, excluding NYHA classification. ^c Unstratified. ^d IC0/1 = PD-L1 IC < 5%. ^e ORR for cis-ineligible IC0/1 pts based on n = 139 in each arm.